Robust Tests for Change in Intercept and Slope in Linear Regression Models with Application to Manager Performance in the Mutual Fund Industry

Financial as well as economic theory have developed models which can be used to evaluate performance of mutual funds and that can assist regulators in monitoring performance of financial markets. These models make use of linear regressions models to provide this information. In particular, interest lies with the intercept parameter of these models and whether it is time varying. In applications to mutual fund performance, time varying alpha indicates a change in manager’s stock selecting abilities. In applications to regulation of financial markets, time-varying alpha indicates potential changes to equity returns based on undisclosed information. These are importance concerns and emphasize the need to have accurate methods to disentangle changes in intercept from slope in these models. Here, a novel bivariate statistic is developed that can be used for this purpose. It has many attractive features. For example, the use of weight functions improves its power for discrete changes in intercept/slope that occur late/early in the sample, allows intercept/slope to change at different dates, allows for control of global error rates; and avoids trimming.